Research On Data Center Virtualization Connected With Optical Networks

In recent years,the diversification of customer demand and the explosive growth of data traffic have made it difficult for the traditional WDM network architecture and transmission technology to meet the current network environment.Therefore,to better support today’s network business demands,network virtualization and elastic optical inter-datacenter networks have emerged.Network virtualization technology solves the problem of network rigidity by allowing multiple independent coexisting heterogeneous virtual networks to share the underlying physical resources,thereby realizing the dynamic allocation and management of resources in the network,and greatly increasing the utilization of network resources.Elastic optical inter-datacenter networks,with multiple advantages of high speed,variable granulity,high scalability,etc.,can well meet the demands of next generation transmission network.This paper investigates the virtualization problem in elastic optical inter-datacenter networks.After giving the problem description,a mathematical model is established,and two virtualization mapping algorithms,HAD-CDC and VT-VMPA are proposed.The HAD-CDC algorithm is used to solve the virtualization mapping problem in VNM scenario,and it is a two-stage heuristic algorithm that divides the mapping process into node and link mapping.The optional data center technology is used in the node mapping phase,where node weights are calculated jointly by the shortest hop distance and distance adaptive modulation bits,so that the physical node with the lowest weight carries the virtual node with the highest CPU demand,and the shortest path algorithm is used in the link mapping phase.Simulation results show that,comparing with the HDA-RDC,the H-CDC and the H-RDC,the HAD-CDC algorithm not only effectively reduces the network resource consumption and blocking rate but also greatly improves the timeaverage gain.The VT-VMPA algorithm is used to solve the virtualization mapping problem in VDCM scenario,where the mapping objective is to minimise the amount of inter-cluster traffic as well as to increase the network mapping rate.In the mapping process,the core virtual machine is firstly searched according to the principle of maximizing the business volume,then the virtual machines that meet the resource constraints and are connected to the core virtual machine but not mapped are continuously merged into clusters according to the descending order of the business volume.By constantly searching for core virtual machines to form different clusters,all virtual machines are formed into clusters,and the mapping is completed by hop distance adaptation and shortest path rules.Simulation results show that the VT-VMPA algorithm has better load balancing,lower network blocking rate,and higher time-average gain compared to the HDA-VMPA algorithm and IC-VMPA algorithm,in addition to consuming less network bandwidth resources.